Premium
More than a mirage: A large‐scale assessment centre with more dimension variance than exercise variance
Author(s) -
Guenole Nigel,
Chernyshenko Oleksandr S.,
Stark Stephen,
Cockerill Tony,
Drasgow Fritz
Publication year - 2013
Publication title -
journal of occupational and organizational psychology
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 2.257
H-Index - 114
eISSN - 2044-8325
pISSN - 0963-1798
DOI - 10.1111/j.2044-8325.2012.02063.x
Subject(s) - variance (accounting) , dimension (graph theory) , construct (python library) , scale (ratio) , psychology , explained variation , applied psychology , construct validity , social psychology , psychometrics , statistics , clinical psychology , computer science , mathematics , geography , accounting , cartography , pure mathematics , business , programming language
Assessment centres ( AC s) are widely recognized to be among the best tools for assessing and developing management talent (Assessment Centres and Global Talent Management, 2011, Gower: London). Yet, the current consensus about the construct validity of AC s is that exercises rather than dimensions explain the majority of variance in ratings. Because much of the data on which these conclusions are based are now old, it is worthwhile to periodically re‐examine this issue to see whether well‐implemented designs produce better measurements of dimensions. We present results from 1,205 executive‐level leaders from F ortune 500 firms across E urope and N orth A merica, who participated in developmental AC s that use modern design principles where assessors were formally examined to ensure they had a common frame or reference. Our results showed that while dimensions and exercises mutually determined ratings, more variation was owing to dimensions. Practitioner Points Modern assessment centre designs can yield dimension scores that reflect dimensions more than exercises; The psychometric characteristics of operational ACs should be regularly monitored as the characteristics of any single AC may differ widely from meta‐analytic averages.